ImageBind

lastmod 2023-05-25created 2023-05-253 edits

ImageBind One Embedding Space to Bind Them All

Meta's multimodal AI model

Features

• Binds data from six modalities at once without supervision, for better analysis of diverse kinds of information
  • Image, video, audio, text, depth, thermal, IMU (inertial measurement unit)
  • Enables tasks such as generating images from sound, or retrieving information by combining audio/video/image/depth data
• Excellent zero-shot / few-shot recognition performance

See also

• Deep learning
• ONE-PEACE - A general representation model for unlimited modalities
• Zero-shot learning
• Few-shot learning

Favorite site

• ImageBind by Meta AI
• facebookresearch/ImageBind